JPS60153567A - Method for extracting area in printed document picture - Google Patents

Method for extracting area in printed document picture

Info

Publication number
JPS60153567A
JPS60153567A JP59009525A JP952584A JPS60153567A JP S60153567 A JPS60153567 A JP S60153567A JP 59009525 A JP59009525 A JP 59009525A JP 952584 A JP952584 A JP 952584A JP S60153567 A JPS60153567 A JP S60153567A
Authority
JP
Japan
Prior art keywords
dimensional fourier
area
character
fourier transformation
fourier transform
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
JP59009525A
Other languages
Japanese (ja)
Inventor
Masahiko Hase
雅彦 長谷
Hiroyuki Hoshino
星野 坦之
Akihiro Shimizu
明宏 清水
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nippon Telegraph and Telephone Corp
Original Assignee
Nippon Telegraph and Telephone Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nippon Telegraph and Telephone Corp filed Critical Nippon Telegraph and Telephone Corp
Priority to JP59009525A priority Critical patent/JPS60153567A/en
Publication of JPS60153567A publication Critical patent/JPS60153567A/en
Pending legal-status Critical Current

Links

Landscapes

  • Image Analysis (AREA)

Abstract

PURPOSE:To shorten a processing time by extracting a peak point corresponding to a character pitch by using one-dimensional Fourier transformation and comparing the size of peaks to extract a character area and a graphic area. CONSTITUTION:The data of an original picture 1 are detected by a picture information detecting part 2 and binary-coded by a binary-coding processing part 3. The one-dimensional Fourier transformation of the x-direction is executed by an one-dimensional Fourier transformation processing part 4 and a pattern area is extracted on the basis of the change of peak point value corresponding to the pitch of a character string. Then, the one-dimensional Fourier transformation of the y-direction is executed again by the processing part 4 and the pattern area is extracted on the basis of the change of the peak point value. Each data of area extraction are stored in an information storage part 5. Thus, the extraction of the character and pattern areas by using the one-dimensional Fourier transformation makes it possible to attain the processing only by the one-dimensional Fourier transformation and shorten the processing time.

Description

【発明の詳細な説明】 〔発明の技術分野〕 この発明は、既存の本や印刷文書中の情報を自動的に入
力する方法において、印刷文書画像中の文字領域と図形
領域の領域抽出を行う方法に関するものである。
[Detailed Description of the Invention] [Technical Field of the Invention] The present invention provides a method for automatically inputting information in an existing book or printed document, in which character areas and graphic areas are extracted from a printed document image. It is related to.

〔従来技術〕[Prior art]

従来の文字1図形領域を抽出する方法としては、第1図
に示すように黒/白両画素のランレングスvllべ、そ
の長さの違いにより各領域を抽出する方法、あるいは第
2図に示すようK、一定領域内の文書の濃度分布め違い
により文字1図形領域を抽出する方法、または゛第3図
に示すよ5K、画素間の近接線密度を用い各領域を抽出
する方法等があった。
Conventional methods for extracting character 1 graphic areas include a method of extracting each area based on the difference in run length of both black and white pixels, as shown in Figure 1, or a method of extracting each area based on the difference in length, as shown in Figure 2. There are two methods, such as a method of extracting a single graphic area of a character based on the difference in the density distribution of a document within a certain area, or a method of extracting each area using the proximity line density between pixels as shown in Figure 3. Ta.

すなわち、第1図〜第3図において、1は印刷文書画像
である原画像、Xは副走査方向、yは主走査方向を示し
、(イ)は文字領域、(ロ)は図形領域である。
That is, in FIGS. 1 to 3, 1 is an original image that is a printed document image, X is a sub-scanning direction, y is a main scanning direction, (a) is a character area, and (b) is a graphic area. .

IIL1図では、例えば図形領域(ロ)を抽出するのに
白のランレングスに着目丁れば、図形領域←)の方が、
文字領域沿より長いことから真領域(イ)、←)の区別
を行うことかできる。
In Figure IIL1, for example, if we focus on the white run length to extract the figure area (b), the figure area ←) is better.
Since it is longer than the character area, the true area (a) and ←) can be distinguished.

第2図では、ある大きさの枠4−→をとって、その内部
の濃度分布tみると、図形領域(ロ)の方が文字領域ピ
)より薄いことから両領域(イ)、(ロ)の区別を行う
ことかできる。
In Figure 2, if we take a frame 4-→ of a certain size and look at the density distribution t inside it, we find that the graphic area (b) is thinner than the character area (pi), so both areas (a) and (ro) ) can be distinguished.

第3図では、ある画素について、矢印に)K示すように
上下、左右の4方向の近接画素の有無から近接線密度を
め、この差から文字領域(イ)と図形驚舅←)l抽出す
るものである。
In Figure 3, for a certain pixel, the adjacent line density is calculated from the presence or absence of adjacent pixels in four directions (up, down, left and right) as shown by the arrows), and from this difference, the character area (a) and the figure surprise are extracted. It is something to do.

しかし、以上3つの方法はいずれも画素間の情報に着目
したものであり、実際に計算機で演算を行った場合は処
理時間はミニコンピユータを用いて数時間におよぶとい
う欠点があった。
However, all of the above three methods focus on information between pixels, and have the disadvantage that when the calculations are actually performed on a computer, the processing time is several hours using a minicomputer.

〔発明の概要〕[Summary of the invention]

この発明は、これらの欠点を解決するため、文書画像l
ラインごとの濃度情報を一次元フーリエ変換を行い、そ
のフーリエ変換された情報より文字ピッチに対応し工ビ
ーク点l抽出し、そのピーク点の変換強度の相対値の違
いにより文子領域と図形領域を切り分けて領域抽出を行
うものである。
In order to solve these drawbacks, this invention
The density information for each line is subjected to one-dimensional Fourier transform, and the peak points corresponding to the character pitch are extracted from the Fourier-transformed information, and the text area and graphic area are determined based on the difference in the relative value of the conversion intensity of the peak point. This is to separate the area and extract the area.

以下、図面についてこの発明を詳l!5VC説明する。This invention will be explained in detail with reference to the drawings below! 5VC will be explained.

〔発明の実施例〕[Embodiments of the invention]

はじめに、この発明の原理について説明し、次いで実施
例について述べる。
First, the principle of this invention will be explained, and then examples will be described.

第4図に示されるような印刷文書画像の濃度分布−g 
f (1,7)とすると、X方向に対するlラインに対
する7−リエ変換は、次のような式で表てことかできる
Density distribution of a printed document image as shown in FIG.
Assuming f (1, 7), the 7-lier transform for the l line in the X direction can be expressed by the following equation.

F (u) = f:二。f (x、y)exp(−j
2πux)dxここでUは、空間周波数である。
F (u) = f: two. f (x,y)exp(-j
2πux)dx where U is the spatial frequency.

ディスクリートな形で表現すると、次のような形で表て
ことができる。
When expressed in a discrete form, it can be expressed in the following form.

ここでNは、対象とする所定の大きさ内の副走査方向の
画素数である。
Here, N is the number of pixels in the sub-scanning direction within a predetermined target size.

一般に、既存の本や原稿の中の文字列は周期性なもつた
めKF(u)は、その周期性に対応した空間周波数Uの
所にピーク点が生じる(第4図、第5図参照)。ピーク
点の空間周波数Uの位置は、文字の周期に対応したもの
である。
In general, character strings in existing books and manuscripts have periodicity, so KF(u) has a peak point at a spatial frequency U corresponding to the periodicity (see FIGS. 4 and 5). The position of the spatial frequency U of the peak point corresponds to the period of the character.

第5図、第6図に1ラインの画像情報tプーリ5−変換
した結果を示す。第5図のようK、文字がある一定ピッ
チで1ライン全部に存在する第4図のAfgAVc沿っ
て走査した場合は、ピーク点の相対値は大きいが、文字
が1ラインの半分程度までしか存在しない第4図のB@
に沿って走査した場合での一次元フーリエ変換のピーク
点の値は、lライン全部に文字が存在する場合に比較し
て減少することになる(第6図参照)。しかし、ピーク
点が存在する空間周波数Uの位置は、文字σ)周期が同
じであることから変化しない。
5 and 6 show the results of one line of image information t-pulley 5 conversion. When scanning along AfgAVc in Figure 4, where K and characters exist on the entire line at a certain pitch as shown in Figure 5, the relative value of the peak point is large, but the characters exist only up to about half of one line. No, B in Figure 4
The value of the peak point of the one-dimensional Fourier transform in the case of scanning along the line is reduced compared to the case where characters are present on all l lines (see FIG. 6). However, the position of the spatial frequency U where the peak point exists does not change because the letter σ) period is the same.

つまり、ピーク点が存在する空間周波数Uの位置での変
換強度の値を比較することにより、X方向での文字領域
と図形領域ン判別できる結果となる。すなわち、第7図
の■、■の領域を判別することは可能となる。
That is, by comparing the values of the conversion intensities at the position of the spatial frequency U where the peak point exists, it is possible to distinguish between character areas and graphic areas in the X direction. In other words, it becomes possible to discriminate between the areas .largecircle. and .largecircle. in FIG.

次に11問題となるのか■の領域内での文字5図形領域
(イ)、(ロ)の抽出方法であり、以下にその方法につ
いて述べる。
Next, problem 11 is a method for extracting the character 5 graphic areas (a) and (b) within the area marked ■, and this method will be described below.

■の領域内のX方向の一次元フーリエ変換を行い、その
結果の文字周期に対応したピーク点の位置での変換強度
の大きさによって両領域(イ)、(ロ)の切り出しを行
うこととする。その概念図を第8図に示す。
Perform a one-dimensional Fourier transform in the X direction within the region (ii), and cut out both regions (a) and (b) based on the magnitude of the transform intensity at the position of the peak point corresponding to the resulting character period. do. A conceptual diagram is shown in FIG.

第8図において、■の部分なX方向に一次元フーy工変
換した結果ビ第9図(a) VCボす。
In Fig. 8, the result of one-dimensional foo-y transform in the X direction of the part marked ■ is shown in Fig. 9 (a).

第9図(a)の−次元フーリエ変換結果には、X方向の
文字周期に対応したピーク点が存在する。
In the -dimensional Fourier transform result shown in FIG. 9(a), there is a peak point corresponding to the character period in the X direction.

また、第8図の■の部分を一次元フーリエ変換した結果
を第9図(b) K示す。
Furthermore, the result of one-dimensional Fourier transformation of the part marked ■ in FIG. 8 is shown in FIG. 9(b)K.

第9図(b)K示す一次元フーリエ変換結果には、原画
像にX方向の周期性が存在しないためピーク点が存在し
ない。つまり、ピーク点の有無外よって文字領域と図形
領域の領域抽出が可能となる。
In the one-dimensional Fourier transform result shown in FIG. 9(b)K, there is no peak point because there is no periodicity in the X direction in the original image. In other words, character areas and graphic areas can be extracted based on the presence or absence of peak points.

以下に具体的な対象例について説明する。Specific target examples will be explained below.

第1O図に処理対象画像(512X512画素)の−例
を示す。次に、原画像を2値化しX方向の第10図に示
す線の位置でX方向に一次元フーリ工変換した結果をm
l1図に示す。第11図に示すようK、文字の周期に対
応するピーク点が検出できる。つまり、1行の文字数に
対応した空間周波数Uの所にピーク点が存在する。
FIG. 1O shows an example of an image to be processed (512×512 pixels). Next, the original image is binarized and the result of one-dimensional Fourier transform in the X direction at the position of the line shown in Figure 10 in the X direction is m
It is shown in Figure l1. As shown in FIG. 11, a peak point corresponding to the period of the character K can be detected. In other words, a peak point exists at a spatial frequency U corresponding to the number of characters in one line.

次k、第12図にy方向の位置に対する各X方向の一次
元フーリエ変換の文字周期に対応するピーク点(この場
合242の変換強度の大きさt示す。なお、Thは変換
強度のしきい値を示す。
Next, Fig. 12 shows the peak point corresponding to the character period of the one-dimensional Fourier transform in each Show value.

第12図より明白なようK、文字かX方向に1行丁べて
存在する文字領域ピ)の場合にはビ り点の値か高く、
文字数が手分しか存在しない図形領域(ロ)の場合には
変換強度の値は小さくなり、明白に文字領域(イ)と図
形領域(ロ)の判別が可能である。
As is clear from Figure 12, in the case of K, a character area (P) that exists in one line in the X direction, the value of the beat point is high;
In the case of a graphic area (b) in which there are only as many characters as hands, the value of the conversion strength is small, and it is possible to clearly distinguish between the character area (a) and the graphic area (b).

次に、前述したように、同じよ5Ky方向に対する一次
元フーリエ変換を行うことにより、X方向の位置座標を
検出てることが可能である。
Next, as described above, by similarly performing one-dimensional Fourier transformation in the 5Ky direction, it is possible to detect the position coordinates in the X direction.

次に、この発明の一実施例について第13図のブロック
図と、第14図の処理ツー−により説明する。なお、第
14図中の■〜■は各ステップを示す。
Next, an embodiment of the present invention will be described with reference to the block diagram in FIG. 13 and the processing tool in FIG. 14. Note that ■ to ■ in FIG. 14 indicate each step.

第13図において、1は原画像であり、2は画像情報検
出S(テイテクタ、s)、3は2値化処理部、4は一次
元フーリエ変換処理部、5は情報蓄積部、6は抽出され
た結果を表示する画像情報表示部、Tは情報制御部、8
はプログラム等を格納するメモリ部、9は共通バスであ
る。
In Fig. 13, 1 is the original image, 2 is the image information detection S (teitor), 3 is the binarization processing section, 4 is the one-dimensional Fourier transform processing section, 5 is the information storage section, and 6 is the extraction section. an image information display section that displays the results, T is an information control section; 8
9 is a memory section for storing programs, etc., and 9 is a common bus.

原画像1のデータは画像情報検出′m2により検出さt
t、2値化処理部3より2値化されるの。次K、−次元
フーリエ変換処理部4でX方向の一次元フーリエ変換処
理が行われ■、領域抽出が行われ文字列のピッチに対応
したピーク点の値の変化より図形領域の抽出を行う■。
The data of original image 1 is detected by image information detection 'm2.
t, it is binarized by the binarization processing unit 3. Next, a one-dimensional Fourier transform process in the X direction is performed in the K, -dimensional Fourier transform processing unit 4, and area extraction is performed, and a graphic area is extracted based on the change in the value of the peak point corresponding to the pitch of the character string. .

次に、再び一次元フーリエ変換処理部4でy方向の一次
元フーリ工変換が行わj■、ピーク点の値の変化より図
形領域の抽出が行われ■、領域抽出された各デ、−夕は
情報蓄積部SK蓄積される。
Next, the one-dimensional Fourier transform in the y direction is again performed in the one-dimensional Fourier transform processing unit 4, and the graphic region is extracted from the change in the value of the peak point. is stored in the information storage section SK.

なお、上記実施例における原画像1は手書による印刷文
書画像でも折目があるか、あるいはきらんと揃えて書い
てあればこの発明を適用することができる。
Note that the present invention can be applied to the original image 1 in the above embodiment even if it is a handwritten printed document image as long as the original image 1 has folds or is written in straight lines.

〔発明の効果〕〔Effect of the invention〕

以上説明したようk、この発明は、印刷文書の文字の周
期性に着目し、−次元フーリエ変換を利用して文字ピッ
チに対応したピーク点を抽出し、ピークの大きさt比較
することにより文字領域と図形領域の位置を抽出するよ
うKしたので、−次元フーリエ変換のみで処理できるk
め従来の方法よりも地理時間が短くてすむ。
As explained above, this invention focuses on the periodicity of characters in printed documents, extracts peak points corresponding to the character pitch using -dimensional Fourier transform, and compares the peak sizes t to determine the character pitch. Since we have set K to extract the positions of regions and figure regions, we can process K using only -dimensional Fourier transformation.
Therefore, the geographical time required is shorter than that of conventional methods.

また、地理か1ライン単位で行うことができるので、フ
ァクス等のライン単位で情報入力する機器においても適
用できる利点かある。
Furthermore, since the geographical information can be performed on a line-by-line basis, it has the advantage that it can also be applied to devices such as fax machines that input information on a line-by-line basis.

【図面の簡単な説明】[Brief explanation of drawings]

第1図はランレングスを用いた領域抽出法の説明図、第
2図は画像の一定領域内の濃度を用いた領域抽出法の説
明図、第3図は任意の点で次の黒点まで距離の加算によ
る領域抽出法の説萌図:第4図は原画像例およびフーリ
エ変換を行うエリアの説明図、第5図は第4図のA@に
沿つに走f、を7−リエ変換した場合のフーリエ変換結
果を示す図、第6図は第4図のB IIAK’沿った走
査tフーリエ変換した場合のフーリエ変換結果を示す図
、第れる領域の説明図、第9図(a)、(b)は第8図
の■の部分の一次元フーリエ変換結果と、第8図の■の
部分の一次元フーリエ変換結果をそれぞれ示す図、第1
0図は処理画像の一例を示す図、第11図は一次元フー
リエ変換処理結果を示す図、第12図は一次元フーリエ
変換結果のピーク点の大きさt示す図、第13図はこの
発明の一実施例を示す □ゾ冒ツク図、第14図は処理
フローを示す図である。 図中、1は原画像、2は画像情報検出部、3は2値化処
理部、4は一次元フーリエ変換処理部、5は情報蓄積部
、6は画像情報表示部、Tμ情報制御部、8はメモリ部
、9は共通パスである。 第1図 第3図 第4図 第5図 第6図 第7図 第8図 第9図 (b) 第10図
Figure 1 is an illustration of the area extraction method using run length, Figure 2 is an illustration of the area extraction method using density within a certain area of the image, and Figure 3 is the distance from any point to the next black point. An illustration of the area extraction method by addition of: Figure 4 is an explanatory diagram of an example of the original image and the area to be Fourier transformed, Figure 5 is the 7-lier transformation of f along A@ in Figure 4 Figure 6 is a diagram showing the results of Fourier transform when scanning along t-Fourier transform is performed along B IIAK' in Figure 4. , (b) is a diagram showing the one-dimensional Fourier transform result of the part marked ■ in Fig. 8, and the one-dimensional Fourier transform result of the part marked ■ in Fig. 8, respectively.
Figure 0 shows an example of a processed image, Figure 11 shows the results of one-dimensional Fourier transform, Figure 12 shows the size t of the peak point of the result of one-dimensional Fourier transform, and Figure 13 shows the results of this invention. 14 is a diagram showing a processing flow. In the figure, 1 is an original image, 2 is an image information detection section, 3 is a binarization processing section, 4 is a one-dimensional Fourier transform processing section, 5 is an information storage section, 6 is an image information display section, a Tμ information control section, 8 is a memory section, and 9 is a common path. Figure 1 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 (b) Figure 10

Claims (2)

【特許請求の範囲】[Claims] (1) 印刷文書画像の文字領域と図形−域を切り分け
る方法において、地理すべき文書画像の1ラインごとの
一次元フーリエ変換結果の文字ピッチに対応するピーク
を検出し、そのピークの相対値の違いKより文字領域と
図形領域との抽出を行うことt特徴とする印刷文書画像
の領域抽出方法。
(1) In the method of separating the character area and graphic area of a printed document image, a peak corresponding to the character pitch of the one-dimensional Fourier transform result for each line of the document image to be mapped is detected, and the relative value of the peak is calculated. A method for extracting regions of printed document images characterized by extracting character regions and graphic regions based on the difference K.
(2) 文書画像の1ラインごとの一次元フニリエ変換
を行う場合に、文書画像をあるしきい値で2値化を行い
、その後に画素ごとの論理和演算を行い、そjから一次
元フーリエ変換を行うことを特徴とする特許請求の範囲
第(llJA記載の印刷文書画画像の領域抽出方法。
(2) When performing a one-dimensional Fourier transform for each line of a document image, the document image is binarized using a certain threshold value, then a logical OR operation is performed for each pixel, and then a one-dimensional Fourier transform is performed for each line. A method for extracting a region of a printed document image according to claim 1 (JA), characterized in that a conversion is performed.
JP59009525A 1984-01-24 1984-01-24 Method for extracting area in printed document picture Pending JPS60153567A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP59009525A JPS60153567A (en) 1984-01-24 1984-01-24 Method for extracting area in printed document picture

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP59009525A JPS60153567A (en) 1984-01-24 1984-01-24 Method for extracting area in printed document picture

Publications (1)

Publication Number Publication Date
JPS60153567A true JPS60153567A (en) 1985-08-13

Family

ID=11722682

Family Applications (1)

Application Number Title Priority Date Filing Date
JP59009525A Pending JPS60153567A (en) 1984-01-24 1984-01-24 Method for extracting area in printed document picture

Country Status (1)

Country Link
JP (1) JPS60153567A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01293485A (en) * 1988-05-23 1989-11-27 Toshiba Corp Picture recognizing device
JP2000339460A (en) * 1999-05-26 2000-12-08 Sharp Corp Region of interest setting device and region of interest setting method
JP2003075708A (en) * 2001-09-07 2003-03-12 Sony Corp Lens device for imaging unit, and flexible printed wiring board used for the lens device

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH01293485A (en) * 1988-05-23 1989-11-27 Toshiba Corp Picture recognizing device
JP2000339460A (en) * 1999-05-26 2000-12-08 Sharp Corp Region of interest setting device and region of interest setting method
JP2003075708A (en) * 2001-09-07 2003-03-12 Sony Corp Lens device for imaging unit, and flexible printed wiring board used for the lens device

Similar Documents

Publication Publication Date Title
JP3308032B2 (en) Skew correction method, skew angle detection method, skew correction device, and skew angle detection device
Gebhardt et al. Document authentication using printing technique features and unsupervised anomaly detection
US5915039A (en) Method and means for extracting fixed-pitch characters on noisy images with complex background prior to character recognition
US5075895A (en) Method and apparatus for recognizing table area formed in binary image of document
JPS61267177A (en) Retrieving system for document picture information
JP2002133426A (en) Ruled line extracting device for extracting ruled line from multiple image
Carter et al. Automatic recognition of printed music
JP2014131278A (en) Method of authenticating printed document
JP2019016350A (en) Identification of emphasized text in electronic documents
US9818028B2 (en) Information processing apparatus for obtaining a degree of similarity between elements
JPS60153567A (en) Method for extracting area in printed document picture
JP2008028716A (en) Image processing method and apparatus
Widiarti et al. Preprocessing model of manuscripts in Javanese characters
JP3171626B2 (en) Character recognition processing area / processing condition specification method
JPH0333990A (en) Optical character recognition instrument and method using mask processing
JP3476595B2 (en) Image area division method and image binarization method
JP2019003534A (en) Image processing program, image processing apparatus, and image processing method
Salagar et al. Analysis of PCA usage to detect and correct skew in document images
JP4548062B2 (en) Image processing device
JP2001126027A (en) Number plate recognition system
JP3162451B2 (en) Preprocessing method for character recognition
JPS5949671A (en) Optical character reader
JPH02252078A (en) Method for identifying area of document
JP2020149426A (en) Character recognition device, character recognition method, and character recognition program
Banerjee et al. Hand-drawn line removal from Bangla printed document images